With the rapid development of science and technology, the issue of hotel accommodation security has become increasingly prominent, especially the increased concealment and technicality of peeping devices, which pose a serious threat to travelers’ privacy. However, deep learning algorithms still face multiple challenges in actual deployment, such as low training efficiency, detection accuracy that needs to be improved, and security threats. This paper optimizes the training parameters of deep learning algorithms and enhances their security in order to improve the detection efficiency and accuracy of hotel anti-peeping devices. For the target system, the specific process and methods of security testing are comprehensively planned, the system is scanned for in-depth vulnerabilities, and hacker attacks are simulated to test its defense capabilities. The detection results of the device under different test conditions are recorded, and the effective detection distance of the device is measured in a preset environment. Through multiple tests, the security indicators of the equipment during data transmission are evaluated. Through the design of experiments, this paper verifies the actual effectiveness of the optimized deep learning algorithm in the hotel anti-voyeurism equipment detection task. The hotel anti-voyeurism equipment optimized by the deep learning algorithm has stable detection capabilities at various test locations, and the horizontal detection distance can mostly meet the standard of not less than 5 m. It can cover all corners of common areas of the hotel more comprehensively, and has good detection performance for locations where peeping devices may be hidden, providing a strong guarantee for the privacy and security of hotel venues.